⚡ electrical engineering

Three-year field performance of a residential grid-tied PV prosumer in Romania: measured data, seasonal yield, and statistical characterisation

This study presents a three-year field analysis of a residential grid-tied PV system in Romania, demonstrating consistent seasonal production patterns, complete self-consumption, and a statistically insignificant performance decline attributed to meteorological variability, thereby addressing the scarcity of real-world PV data in Central-Eastern Europe.

Original authors: George-Andrei Beșchea, Beatrice-Georgiana Vuțoiu, Gabriel Năstase

Published 2026-06-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: George-Andrei Beșchea, Beatrice-Georgiana Vuțoiu, Gabriel Năstase

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: A Solar "Fitness Tracker" for a Romanian Home

Imagine a homeowner in Romania who installed solar panels on their roof. Instead of just hoping the panels work, they decided to keep a detailed diary of exactly how much electricity they made every single day for three years (from 2023 to early 2026).

This paper is the report card of that diary. The researchers, who are like "solar detectives," analyzed this data to answer three main questions:

How much power did the system actually make?
Did the system get "sick" or break down over time?
Why did some years produce less power than others?

The Main Findings

1. The "Perfect Match" (Zero Waste, Zero Borrowing)

The most impressive part of this story is how the house uses its power. Usually, solar homes are like a person with a piggy bank: they save extra money (power) when they have too much and borrow from the bank (the grid) when they run out.

But this specific house is different. For 37 months straight, it never borrowed a single penny from the bank, and it never had a penny left over to put in the bank.

The Analogy: Think of it like a chef cooking a meal exactly to the number of hungry guests at the table. If there are four guests, the chef cooks exactly four portions. No food is wasted, and no one goes hungry. This house perfectly matched its cooking (solar power) to its appetite (electricity usage) every single month.

2. The "Seasonal Rollercoaster"

The solar panels don't produce power evenly all year; they follow the sun's schedule.

Summer (The Rollercoaster Peak): In the summer (July–September), the panels were working overtime, producing about 33–37% of the entire year's power in just three months.
Winter (The Quiet Valley): In the winter (October–December), production dropped significantly, sometimes to less than 10% of the summer output.
The Analogy: Imagine a fruit tree. It doesn't grow apples in January. It produces a massive, juicy harvest in late summer, and then goes dormant in winter. This solar system behaves exactly like that tree.

3. The "Mystery of the Missing Power" (2025)

The system had a great year in 2023 and a similar year in 2024. But in 2025, the total power dropped by about 10%. The researchers had to play detective to figure out why.

They considered a few suspects:

Suspect A: The Panels got dirty. (Like a dusty window). Verdict: Unlikely. Even if the panels were dusty, it wouldn't cause such a big drop.
Suspect B: The equipment broke. (Like a flat tire). Verdict: Unlikely. The pattern of power loss didn't look like a broken machine; it looked like a weather pattern.
Suspect C: The Weather changed. (Like a cloudy summer). Verdict: Guilty.

The Conclusion: The drop in 2025 wasn't because the solar panels were failing. It was because the summer of 2025 was just cloudier and rainier than usual in Romania. The winter of 2025 was actually sunnier than previous years, which balanced things out a bit. The system is healthy; the sky just wasn't as cooperative that year.

4. The "Statistical Proof"

To be sure the system wasn't slowly breaking down, the researchers used a special math test (called the Mann-Kendall test).

The Analogy: Imagine watching a runner for three years. If they slow down a little every year, they are getting tired (degrading). If they run fast one year, slow down the next because of rain, and speed up the next because of a tailwind, they are just reacting to the weather.
The Result: The math showed no evidence that the system is getting tired or breaking down. The "slowdown" was just the weather acting up. The shape of the power curve stayed the same every year, proving the system is stable.

What We Don't Know (The Missing Piece)

The paper admits one big missing piece of the puzzle: We don't know the exact size of the solar panels.

The Analogy: It's like knowing a car drove 3,000 miles, but not knowing if it was a tiny Mini Cooper or a big Truck. We can guess the size based on how much fuel it used and the terrain, but we don't have the official registration papers.
The Guess: Based on the math, the researchers estimate the system is likely between 2.7 and 3.0 kW (a standard size for a home). Even without the exact number, they confirmed the system is performing exactly as expected for a house in that part of Europe.

The Bottom Line

This paper proves that:

Romanian solar panels work great and can perfectly match a home's needs without needing to buy or sell electricity to the grid.
Weather is the boss. If a year produces less power, it's usually because the sun was hiding behind clouds, not because the panels are broken.
You can trust the data. Even without fancy extra sensors, just looking at the monthly numbers from the solar app is enough to tell if a system is healthy or just having a "bad weather year."

In short: The solar panels are healthy, the house is running efficiently, and the only thing that changed was the weather.

Technical Summary: Three-Year Field Performance of a Residential Grid-Tied PV Prosumer in Romania

Problem Statement
Published real-world performance datasets for residential photovoltaic (PV) systems in Central-Eastern Europe, and Romania specifically, remain scarce. While simulation tools (e.g., PVsyst, PVGIS) provide long-term yield estimates, they cannot capture site-specific phenomena such as inverter degradation, module soiling, partial shading, or the precise temporal alignment required for self-consumption. Existing literature is dominated by Western and Southern European studies, leaving a gap in empirical monitoring data for the temperate continental climate of Romania. Furthermore, the rapid proliferation of cloud-connected inverters has generated vast amounts of unpublished operational data, necessitating methodologies to extract and validate this information for scientific analysis.

Methodology
This study analyzes 37 months of continuous production data (January 2023 to January 2026) from a residential grid-tied PV installation in Romania. The system operates under a strict 100% self-consumption mode with zero grid import and zero grid export documented throughout the monitoring period.

Data Acquisition: Monthly production totals were manually extracted from screenshots of the Growatt ShinePhone mobile application, which aggregates minute-level data from a ShineWiFi-X module.
Validation: Data integrity was ensured through two levels of validation: internal consistency checks (ensuring self-consumed + exported = generated) and cross-validation against independent annual totals provided by the platform. All 37 monthly data points matched the annual aggregates with zero discrepancy.
Analytical Framework: The study employed three primary analytical procedures:
1. Descriptive Analysis: Monthly and seasonal (quarterly) production profiles were constructed to identify unimodal seasonal patterns and inter-annual variability.
2. Statistical Characterization: Formal inferential analyses included a non-parametric Mann-Kendall trend test to detect monotonic degradation, pairwise Pearson correlation to assess structural conservation of seasonal profiles, and Seasonal-Trend decomposition using Loess (STL) to separate trend, seasonal, and residual components.
3. Capacity Inference: Due to the absence of verified nameplate documentation for the PV array's rated peak power ( $P_{nom}$ ), an inverse sensitivity analysis was conducted. Observed annual yields were compared against PVGIS-SARAH3 climatological expectations for central Romania (approx. 1,050–1,150 kWh/kWp/yr) to infer the most probable installed capacity.

Key Results

Production Yields: The system produced 3,195.0 kWh (2023), 3,171.3 kWh (2024), and 2,883.7 kWh (2025), with a three-year mean of 3,083.3 kWh/year.
Seasonal Distribution: Production follows a strongly unimodal curve typical of fixed-tilt systems at mid-latitudes. Quarter 3 (July–September) contributed 33–37% of annual yield, while Quarters 2 and 3 combined accounted for 68–70% of total production. Quarter 4 (October–December) was the weakest, contributing 12.5–14.7%.
Inter-Annual Variability: A 9.7% year-on-year decline was observed in 2025 relative to the 2023 baseline. Statistical analysis attributes this primarily to meteorological variability rather than technical degradation. The Mann-Kendall test returned $p = 0.376$ , indicating no significant monotonic trend. Pairwise correlations between annual seasonal profiles exceeded $r = 0.88$ , confirming the structural conservation of the production pattern.
Self-Consumption: The system achieved 100% self-consumption (zero import/export) across all 37 months, indicating a precise alignment between the household's load profile and the PV generation profile.
Capacity Inference: Based on the sensitivity analysis against PVGIS expectations, the most probable installed capacity is inferred to be in the range of 2.7 to 3.0 kWp. This yields a specific production of approximately 1,030–1,180 kWh/kWp/yr, consistent with regional climatological norms.

Significance and Claims
The paper claims to provide one of the first peer-reviewed, long-term field monitoring datasets for a Romanian residential prosumer. Its primary contributions are:

Empirical Evidence: It offers field-validated performance data for a temperate continental climate, filling a gap in the literature where data has previously been limited to grey literature or modeled estimates.
Methodological Template: It demonstrates a reproducible workflow for extracting, validating, and statistically analyzing data from proprietary cloud-connected inverter platforms (specifically Growatt) without requiring additional instrumentation.
Operational Insight: It confirms that 100% annual self-consumption is achievable in Romanian residential settings under appropriate sizing conditions, a performance level significantly higher than the ~30% baseline typical for unstoraged residential PV in Europe.
Attribution of Variability: The study provides a quantitative basis for distinguishing between meteorological anomalies and technical degradation, concluding that the observed 2025 production deficit is driven by inter-annual weather patterns (likely reduced summer irradiance) rather than system failure or degradation.

The authors acknowledge limitations, including the lack of a verified $P_{nom}$ (precluding exact calculation of Performance Ratio), the absence of ground-truth irradiance measurements (pyranometer), and the 37-month duration being below the 5–10 year threshold recommended for robust degradation trend detection. Consequently, the capacity inference is presented as a sensitivity-based estimate rather than a direct measurement.